Build an AI Customer Support Summarizer with MCP (Intercom + Zendesk Recipe)

Support teams drown in tickets. This recipe helps you build an AI assistant that pulls conversations from Intercom or Zendesk, summarizes them, and surfaces trends and urgent issues.

You'll be able to ask things like:

• "Summarize all open Intercom conversations from the last 24 hours and flag anything that mentions 'billing' or 'refund'."
• "Give me a daily digest of Zendesk tickets: count by priority, list the top 5 longest-open tickets, and suggest which need escalation."
• "Compare this week's support volume to last week and list the top 3 topics."

What You'll Build

The flow: MCP tools fetch tickets/conversations → AI reads and summarizes → you get a concise report (in chat or exported).

Prerequisites

Step 1: Set Up Intercom or Zendesk MCP

Use an MCP server that wraps the Intercom or Zendesk API (e.g. community Intercom MCP, Zendesk MCP). Configure it with your API key or OAuth credentials so it can list conversations, get ticket details, and optionally update tickets.

Typical tools you'll see: list_conversations, get_conversation, search_tickets, get_ticket, list_agents.

Step 2: Configure in Your MCP Client

Add the server to Claude Desktop or Cursor. For HTTP-based servers, use the server URL and auth header. For STDIO servers, set command and env vars for the API key.

Step 3: Example Prompts

• "Fetch all Zendesk tickets from the last 24 hours. Summarize by priority and list the 3 most urgent. For each, give a one-line summary and the ticket URL."
• "Get open Intercom conversations. Group by topic (billing, technical, onboarding) and tell me how many are in each group and which have been waiting longest."
• "Create a Monday morning digest: compare ticket volume last week vs the week before, top 5 tags, and suggest one process improvement."

Best Practices

• Use read-only API scopes when you only need summaries; add write scope only if you want the AI to update tickets.
• For large teams, run digests on a schedule (e.g. cron + script that calls your AI with a fixed prompt) and post to Slack or email.

Related Recipes

• Meeting Notes → Jira Tasks
• AI GitHub PR Reviewer

Frequently Asked Questions